| 摘要 |
<p>Here, we worked on the speech processing technique wherein we took audio speech as input signal from user and by recognition process, we perceive recorded word. Specifically we concentrated on the amplitude and zero cross of speech signal, and two of the most important and useful terms, namely MFCC and Mahalanobis distance, for the purpose of getting an efficient outcomes. We made a vast database of different words spoken by various speakers, which contains all magnified researched details of manipulated features. Now for real time comparison we took a speech signal from a user and stored it in digital format, then manipulate algorithm features likewise Amplitude and Zero-cross for block as well in sub-block. Store these manipulate result of real time input speech and calculate MFCC of that, after that we take a decision for recognition. If input speech matched with database stored speech that result displayed otherwise it gives message to record again.</p> |
